Despite several attempts to accurately predict duration and cost of projects, simulation models in use are still over simplified and nonrealistic. They often fail to cope with real-life scenarios and uncertainty. In this paper we use the proxel-based simulation method to analyze and predict duration of project schedules exhibiting high uncertainty due to typical on-the-fly human decision behavior. The proxel-based simulation is an approximate simulation method that is more precise than discrete-event simulation. To model uncertainty, we introduce a new type of task, state-dependent (floating) task that supports and demonstrates a high degree of uncertainty and depends on various parameters in the schedule. For example, the probability distribution of the duration of a task can change depending on the team that performs it. Thus, this kind of task can be used to model the frequent re-scheduling in a project. We use software development process to illustrate our approach.